基于知识表示学习的名老中医治疗肺癌用药知识自动化补全研究

王欣宇; 杨涛; 胡孔法

doi:10.14148/j.issn.1672-0482.2023.0972

基于知识表示学习的名老中医治疗肺癌用药知识自动化补全研究

doi: 10.14148/j.issn.1672-0482.2023.0972

王欣宇^1,,
杨涛¹,
胡孔法^{1, 2, 3, ,}

1.
南京中医药大学人工智能与信息技术学院, 江苏南京 210023
2.
江苏省中医药防治肿瘤协同创新中心, 江苏南京 210023
3.
江苏重大健康风险管理与中医药防控政策研究中心, 江苏南京 210023

基金项目:

国家自然科学基金面上项目 82074580

国家重点研发计划 2022YFC3500201

江苏省重点研发计划 BE2022712

江苏省社会科学项目 21TQB001

详细信息

作者简介:
王欣宇, 男, 博士研究生, E-mail：20200996@njucm.edu.cn

通讯作者:
胡孔法, 男, 教授, 主要从事中医药人工智能与大数据分析研究，E-mail：kfhu@njucm.edu.cn

中图分类号: R2-03
计量
- 文章访问数: 142
- HTML全文浏览量: 19
- PDF下载量: 16
- 被引次数: 0
出版历程
- 收稿日期: 2023-04-23
- 网络出版日期: 2023-11-10
- 发布日期: 2023-10-10

An Automated Completion Study of Knowledge for the Treatment of Lung Cancer by Famous TCM Experts Based on Knowledge Representation Learning

WANG Xin-yu^1
,,
YANG Tao¹,
HU Kong-fa^{1, 2, 3
, ,}

1.
School of Artificial Intelligence and Information Technology, Nanjing University of Chinese Medicine, Nanjing 210023, China
2.
Jiangsu Collaborative Innovation Center of Traditional Chinese Medicine in Prevention and Treatment of Tumor, Nanjing 210023, China
3.
Jiangsu Research Center for Major Health Risk Management and TCM Control Policy, Nanjing 210023, China

摘要

摘要: 目的针对名老中医治疗肺癌医案用药知识库构建任务, 实现知识库的自动化补全。方法设计名老中医治疗肺癌用药知识补全方法, 对症状、诊断、中药、舌象和脉象之间的相关性进行分析, 并根据阈值进行划分, 在组成初始知识库的基础上使用CrossE知识表示模型结合负采样技术学习知识库中各实体和关系在向量空间中的嵌入表示, 通过链路预测任务挖掘知识库中的隐藏关系。结果在对医案中各实体进行皮尔逊相关性分析的基础上利用CrossE模型所预测的关系在Hit@1、Hit@3、Hit@5和Hit@10指标上分别达到了16.19%、29.12%、35.85%和47.60%, 在MeanRank指标上达到了13.19。相较TransE、TransR等模型, 有显著提升。结论使用知识补全技术结合中医临床实践可以深入挖掘名老中医治疗肺癌的隐藏知识。
- 知识表示学习 /
- CrossE /
- 名老中医 /
- 肺癌用药 /
- 知识自动化补全
Abstract: OBJECTIVE To build the knowledge base of medical cases for the treatment of lung cancer by famous TCM experts, and to realize the automatic completion of the knowledge base. METHODS A method of completing the drug use of famous old Chinese medicine for the treatment of lung cancer was designed, the correlation between symptoms, diagnosis, traditional Chinese medicine, tongue image and pulse image was analyzed and divided according to the threshold, and the CrossE knowledge representation model was used in conjunction with negative sampling techniques to learn the embedded representation of each entity and relationship in the knowledge base in the vector space, and the link prediction task was carried out to mine the hidden relationships in the knowledge base. RESULTS Based on the Pearson correlation analysis conducted on entities within medical cases, the relationships predicted using the CrossE model achieved the following results for the Hit@1, Hit@3, Hit@5, and Hit@10 metrics: 16.19%, 29.12%, 35. 85%, and 47.60%, respectively. Additionally, the MeanRank metric reached 13.19. Compared to models such as TransE and TransR, there is a significant improvement. CONCLUSION The use of knowledge completion technology in conjunction with clinical practices in TCM can deeply explore the hidden knowledge of famous TCM experts in the treatment of lung cancer.
- knowledge representation learning /
- CrossE /
- famous TCM experts /
- lung cancer medication /
- knowledge automation completion

HTML全文

图 1 研究结构图

Figure 1. Research structure diagram

下载: 全尺寸图片幻灯片

图 2 相关性热力图

Figure 2. Correlation heat map

下载: 全尺寸图片幻灯片

图 3 初始知识库结构示意图

Figure 3. Schematic diagram of the initial knowledge base structure

下载: 全尺寸图片幻灯片

图 4 链路预测结果

注：A.病机“饮停胸胁”对应的症状；B.病机“饮停胸胁”对应的中药

Figure 4. Link prediction results

下载: 全尺寸图片幻灯片

图 5 链路预测结果

Figure 5. Link prediction results

下载: 全尺寸图片幻灯片

表 1 语义关系表

Table 1. Semantic relationship table

关系类型	关系含义	详细描述	阈值
pat_ton	诊断-舌象	相应诊断所易表现出的舌象	0.05
pat_pul	诊断-脉象	相应诊断所易表现出的脉象	0.05
pat_sym	诊断-症状	相应诊断所易表现出的症状	0.05
pat_her	诊断-中药	相应诊断所常使用的中药	0.05
ton_sym	舌象-症状	相应舌象所易表现出的症状	0.05
pul_sym	脉象-症状	相应脉象所易表现出的症状	0.05
sym_her	症状-中药	相应症状所常使用的中药	0.1

下载: 导出CSV

表 2 测试集评估结果

Table 2. Test set evaluation results

模型	MeanRank	Hit@1/%	Hit@3/%	Hit@5/%	Hit@10/%
TransE	17.03	14.92	24.71	30.44	40.57
TransH	25.38	8.48	16.91	22.47	32.34
TransR	23.29	10.07	18.87	24.47	34.58
TransD	13.68	14.30	26.81	34.98	46.74
ConvKB	45.20	6.01	11.78	15.93	22.63
ERMLP	25.41	7.54	15.52	21.03	31.00
CrossE	13.19^*	16.19^*	29.12^*	35.85^*	47.60^*
注: ^*为最优结果。

下载: 导出CSV

表 3 CrossE模型测试结果

Table 3. CrossE model test results

模型	MeanRank	Hit@1/%	Hit@3/%	Hit@5/%	Hit@10/%
未负采样	14.18	15.47	27.25	35.03	45.23
负采样	13.19^*	16.19^*	29.12^*	35.85^*	47.60^*
注: ^*为最优结果。

下载: 导出CSV

参考文献(14)

[1]	赵飞. 高能光子放疗的PET剂量验证研究[D]. 苏州: 苏州大学, 2013. ZHAO F. 3-D PET dose verification of energetic photon radiotherapy[D]. Suzhou: Soochow University, 2013.
[2]	蔡云, 陈远彬, 叶放, 等. 国医大师周仲瑛从癌毒理论辨治肺癌经验述要[J]. 中华中医药杂志, 2020, 35(6): 2879-2882. https://www.cnki.com.cn/Article/CJFDTOTAL-BXYY202006047.htm CAI Y, CHEN YB, YE F, et al. Brief on TCM master ZHOU Zhong-ying's experience in the treatment of lung cancer based on cancerous toxin theory[J]. China J Tradit Chin Med Pharm, 2020, 35(6): 2879-2882. https://www.cnki.com.cn/Article/CJFDTOTAL-BXYY202006047.htm
[3]	杨大伟, 周刚, 卢记仓, 等. 基于知识表示学习的知识图谱补全研究综述[J]. 信息工程大学学报, 2021, 22(5): 558-565. https://www.cnki.com.cn/Article/CJFDTOTAL-XXGC202105007.htm YANG DW, ZHOU G, LU JC, et al. Review of knowledge graph completion based on knowledge representation learning[J]. J Inf Eng Univ, 2021, 22(5): 558-565. https://www.cnki.com.cn/Article/CJFDTOTAL-XXGC202105007.htm
[4]	李杳瑶, 刘华, 孙铜林, 等. 基于数据挖掘探析刘华教授治疗非小细胞肺癌的学术思想[J]. 湖南中医药大学学报, 2023, 43(2): 327-331. https://www.cnki.com.cn/Article/CJFDTOTAL-HNZX202302023.htm LI YY, LIU H, SUN TL, et al. Professor LIU Hua's academic thoughts on the treatment of non-small cell lung cancer by data mining[J]. J Hunan Univ Chin Med, 2023, 43(2): 327-331. https://www.cnki.com.cn/Article/CJFDTOTAL-HNZX202302023.htm
[5]	申刚磊, 胡孔法, 杨涛, 等. 基于复杂网络的名中医诊治肺癌用药规律分析[J]. 世界科学技术-中医药现代化, 2019, 21(3): 501-505. https://www.cnki.com.cn/Article/CJFDTOTAL-SJKX201903028.htm SHEN GL, HU KF, YANG T, et al. Analysis of medication rule of famous Chinese medicine in diagnosis and treatment of lung cancer based on complex network[J]. Mod Tradit Chin Med Mater Med World Sci Technol, 2019, 21(3): 501-505. https://www.cnki.com.cn/Article/CJFDTOTAL-SJKX201903028.htm
[6]	许玉龙, 段涛, 刘方方, 等. 改进贝叶斯统计挖掘名老中医对肺痿的证候分型经验[J]. 世界科学技术-中医药现代化, 2022, 24(3): 961-967. https://www.cnki.com.cn/Article/CJFDTOTAL-SJKX202203010.htm XU YL, DUAN T, LIU FF, et al. Improved Bayesian statistics to research the syndrome classification rules of lung impotence based on famous TCM doctors' experience[J]. Mod Tradit Chin Med Mater Med World Sci Technol, 2022, 24(3): 961-967. https://www.cnki.com.cn/Article/CJFDTOTAL-SJKX202203010.htm
[7]	ZHANG W, PAUDEL B, ZHANG W, et al. Interaction embeddings for prediction and explanation in knowledge graphs[C]//Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining. February 11-15, 2019, Melbourne VIC, Australia. New York: ACM, 2019: 96-104.
[8]	WALDMANN P. On the use of the Pearson correlation coefficient for model evaluation in genome-wide prediction[J]. Frontiers Genet, 2019, 10: 899.
[9]	BORDES A, USUNIER N, GARCIA-DURÁN A, et al. Translating embeddings for modeling multi-relational data[C]//Proceedings of the 26th International Conference on Neural Information Processing Systems-Volume 2. December 5-10, 2013, Lake Tahoe, Nevada. New York: ACM, 2013: 2787-2795.
[10]	WANG Z, ZHANG JW, FENG JL, et al. Knowledge graph embedding by translating on hyperplanes[C]//Proceedings of the Twenty-Eighth AAAI Conference on Artificial Intelligence. July 27-31, 2014, Québec City, Québec, Canada. New York: ACM, 2014: 1112-1119.
[11]	LIN YK, LIU ZY, SUN MS, et al. Learning entity and relation embeddings for knowledge graph completion[C]//Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence. January 25-30, 2015, Austin, Texas. New York: ACM, 2015: 2181-2187.
[12]	JI GL, HE SZ, XU LH, et al. Knowledge graph embedding via dynamic mapping matrix[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Beijing, China. Stroudsburg, PA, USA: Association for Computational Linguistics, 2015: 687-696.
[13]	NGUYEN DQ, NGUYEN TD, NGUYEN DQ, et al. A novel embedding model for knowledge base completion based on convolutional neural network[EB/OL]. [2023-08-01]. https://arxiv.org/abs/1712.02121.
[14]	DONG X, GABRILOVICH E, HEITZ G, et al. Knowledge vault: A web-scale approach to probabilistic knowledge fusion[C]//Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. August 24-27, 2014, New York, New York, USA. New York: ACM, 2014: 601-610.